Knowledge Status and Sampling Strategies to Maximize Cost-Benefit

www.nature.com/scientificreports

OPEN Knowledge status and sampling strategies to maximize cost- beneft ratio of studies in landscape genomics of wild plants Alesandro Souza Santos* & Fernanda Amato Gaiotto

To avoid local extinction due to the changes in their natural ecosystems, introduced by anthropogenic activities, species undergo local adaptation. Landscape genomics approach, through genome– environment association studies, has helped evaluate the local adaptation in natural populations. Landscape genomics, is still a developing discipline, requiring refnement of guidelines in sampling design, especially for studies conducted in the backdrop of stark socioeconomic realities of the rainforest ecologies, which are global biodiversity hotspots. In this study we aimed to devise strategies to improve the cost-beneft ratio of landscape genomics studies by surveying sampling designs and genome sequencing strategies used in existing studies. We conducted meta-analyses to evaluate the importance of sampling designs, in terms of (i) number of populations sampled, (ii) number of individuals sampled per population, (iii) total number of individuals sampled, and (iv) number of SNPs used in diferent studies, in discerning the molecular mechanisms underlying local adaptation of wild plant species. Using the linear mixed efects model, we demonstrated that the total number of individuals sampled and the number of SNPs used, signifcantly infuenced the detection of loci underlying the local adaptation. Thus, based on our fndings, in order to optimize the cost-beneft ratio of landscape genomics studies, we suggest focusing on increasing the total number of individuals sampled and using a targeted (e.g. sequencing capture) Pool-Seq approach and/or a random (e.g. RAD- Seq) Pool-Seq approach to detect SNPs and identify SNPs under selection for a given environmental cline. We also found that the existing molecular evidences are inadequate in predicting the local adaptations to climate change in tropical forest ecosystems.

Anthropogenic activities are transforming natural systems, drastically changing the environmental conditions in a way which poses major threat to global biodiversity1,2. Anthropogenic modifcations can lead to reduction and fragmentation of natural environments, or to changes in climatic conditions3–5. Species respond to such changes in the natural habitat, by (i) phenotypic plasticity, (ii) migrating from their natural habitats, in search of conditions, ft for survival and having ample resources, or (iii) adapting to the new environment to avoid local extinction (local adaptation)4,6–8. A species is said to exhibit local adaptation, when individuals on an average have a superior ftness in their home environment, compared to a transplanted individual8–11. Terefore, local adaptation is driven by the action of natural selection, which acts on the individual’s phenotypes, determining the characteristics that will be favored under certain environmental conditions9,12,13. Historically, local adaptation has been studied either through trans- location experiments (between environments) or trials in the greenhouse under controlled environmental con- ditions8,14,15. Te drawbacks of these two approaches include the requirement of ample fnancial resources and time, which is generally scarce for studies involving long-lived species such as trees14,15. Landscape genomics, a study that identifes genetic variations that confer local adaptation, is used to remedy these limitations16. In this approach, signifcant diferences in allele frequency between populations of the target species, indicates that individuals in the population are experiencing selection pressure; possibly in response to change in some environmental factor, such as changes in soil type, radiation, water stress, and temperature17–20. Landscape genomics studies analyze frequency distribution changes in molecular markers, such as single nucleotide polymorphism

Applied Ecology & Conservation Lab, Santa Cruz State University - Rodovia Ilhéus-Itabuna, km 16, Ilheus, BA, ZIP Code: 45662-901, Brazil. *email: [email protected]

Scientific Reports | (2020)10:3706 | https://doi.org/10.1038/s41598-020-60788-8 1 www.nature.com/scientificreports/ www.nature.com/scientificreports

(SNPs)14,21 in relation to given environmental factors. Te SNPs are commonly used in wild local adaptation studies because their location and functional annotations are known and they are widely distributed throughout the genome22,23. Additionally, the advent of high-throughput sequencing technology (HTS) has made the sequencing of millions of SNPs from the genome, possible, at moderate cost and it is not time intensive24,25. Tus, through the use of landscape genomics, it is now possible to fnd correlations between genomic regions and the variable environmental characteristics15,21. Terefore, this approach is used to pinpoint the environmental change in nature, that afects ecology (e.g.: climatic changes) of a species and can infuence its adaptive genetic potential26–28. Tus, landscape genomics approach is useful in predicting the responses of species to environmental heterogeneity and variable landscape factors29. In landscape genomics studies, outlier loci method has been used in tandem with genome–environment association (GEA) method, to evaluate local adaptation in natural populations21. In the outlier loci method, changes in among-population frequency distribution of given allele(s)/loci, signifcantly diferent from that seen in the absence of natural selection, are identifed, and these alleles/loci are considered to be under natural selection pres- sure15,21. While, in the GEA analysis, occurrence of high correlation between the allele frequencies with one or more environmental variables is considered an indicator of local adaptation10,30. However, as the outlier method does not specify the environmental forces at work on the locus under selection, studies using landscape genomics approach use the GEA results31,32 to fx the causative environmental factor. For this reason, in this study we tried to systematize and discuss the results obtained from GEA in wild populations, sampled in situ. Considering that landscape genomics is a relatively new discipline, it requires refnement within the scope of sampling design, for GEA studies33–37. Studies, based on simulations, have predicted that the number of SNP markers used and the number of populations sampled infuence the inferences from landscape genomics analy- sis38,39. However, a recent review article, evaluated the impacts of sampling design on inferences from empirical studies on landscape genomics and it does not take into account the particularities of tropical regions37. In a way, our approach complements that of Ahrens et al., as they focused on the limitations of the diferent techniques, the lack of standardization and non-availability of information in conducting these studies. However, unlike others, our study, by reporting on empirically observed patterns in landscape genomics, aims to shape sampling design strategies for improving the cost-beneft ratio of future works, conducted in laboratories using population genomics for the frst time. Guidelines put forth in this study will be helpful, in particular, in subsidizing the costs of studies being carried out in the stark socioeconomic realities of tropical regions, which are global biodiversity hotspots under grave anthropogenic threat. Methodology To perform this work, all research articles, published to date (at the time of writing, September 2018), which evaluated local adaptations in populations of wild plants were pooled. Te papers were analyzed for (i) number of populations sampled, (ii) number of individuals sampled per population, (iii) total number of individuals sampled, and (iv) number of SNPs used. Care was taken to ensure that all data were obtained in empirical studies assessing local adaptation in wild plant populations. Te Scopus (https://www.scopus.com) and Google Scholar (http://scholar.google.com.br/) databases were searched for title, abstract, and articles using the following key- words: environmental variables and SNPs, landscape genetics and SNPs, spatial analysis and SNPs, landscape genomics and SNPs, population genomics and SNPs, adaptive genetic variation, and local adaptation and SNPs. Search results were fltered to remove papers and reviews from clinical, biomedical, veterinary, and immunology areas. Ten, a second flter was applied to eliminate the articles containing the following words: animal, fsh, ecology of freshwater, marine biology, entomology, and zoology. Finally, papers using simulations, or involving exotic and crop species in plantations, or using greenhouse experiments were removed and the remaining 35 empirical studies on in situ wild plants were selected for the present work. Using this data set and ArcGIS (10.2), a map was created depicting the distribution of the localities where researches were performed, in the diferent terrestrial biomes (Fig. 1). A double check was done at this stage to ensure if all the 35 papers chosen for this study, had evaluated in situ local adaptation in wild plant populations, using SNPs markers. Te following information from the papers: (1) authors names; (2) publication year, (3) country of study and geographical coordinates; (4) botanical family studied; (5) species; (6) total number of individuals; (7) number of populations; (8) number of SNPs used; (9) number of SNPs under selection; and (10) method used to generate the SNPs data, was compiled. For this study, we considered as SNPs under selection, the SNPs from the pooled data that had a signifcant association with a geoclimatic variable, such as temperature, precipitation, latitude, longitude, elevation, evapo- transpiration, and drought. Tey were used as a proxy to detect the potential for natural selection in wild plant populations. Te number of SNPs under selection could be infuenced by the methods employed to analyze the correlation between allelic frequency and environmental variables38, therefore most articles in the literature make inferences from such studies, using conservative criteria34,35,40. It is known that each method has diferent types of limitations and caveats (e.g., vary in the rates of false positives or of false negatives), which can cause errors in the estimates and consequently in the estimate of the number of SNPs under selection38. For this reason, we chose to be conservative, using only the results obtained either simultaneously by at least two methods of analysis or strictly controlled for rates of false positives or false negatives (e.g., studies that controlled the population structure in the analyses). Tus, although limitations of the diferent methodologies used were not fully circumvented, we believe that the results reported in our mini-review, portrays the fndings of the literature, in a way seeking to minimize, as much as possible, the biases inherent to the diferent analysis. Te methods used to generate the SNPs data were subdivided into Random, Random Pool-Seq, Targeted, and Targeted-Pool-Seq categories. Te Random category included studies that used random regions of DNA and individualized sequencing for the preparation of libraries. Te Random-Pool-Seq category is distinct from Random, in that the data is sequenced with the Pool-Seq technique (Pool-Seq, an equimolar amount of DNA is taken from each individual from a population and pooled for sequencing). In the Targeted category, we included studies that

Scientific Reports | (2020)10:3706 | https://doi.org/10.1038/s41598-020-60788-8 2 www.nature.com/scientificreports/ www.nature.com/scientificreports

Figure 1. Geographical, ecological and species distribution found in the 35 research articles, included in this study, which evaluated local adaptation in wild plant populations in their natural occurrence area using SNPs markers, are shown; (A) Percentage of studies published per continent; (B) Percentage of studies distributed per plant family; (C) Te points distributed on the map indicating the sampling of the studies in the diferent terrestrial biomes.

used only specifc gene regions or expressed sequence tags with individualized sequencing. Te Targeted Pool-Seq category difered from Targeted, such that the data is sequenced by Pool-Seq technique mentioned above. Infuence of sample size on number of SNPs under selection was inferred by determining the average of the individual number of samples from each population which was arrived at by dividing the total number of individuals and the number of populations in each study. Additionally, the percentage of SNPs under selection was calculated to estimate the number of genes under natural selection in the genome. Ten, a descriptive statistical analysis (minimum, maximum, mean, and standard) was performed with all the variables in R program (http:// www.r-project.org/). To evaluate the infuence of each variable on the number of SNPs under selection, the linear mixed efects model of the lme4 package of R program41 was used. Te variables in this study, such as the number of populations, number of individuals, average number of individuals sampled per population, and number of SNPs used were investigated to determine their infuence on the number of SNPs under selection and, consequently, the detection of natural selection in plant species. As this type of model requires the fulfllment of the assumptions of normality and homoscedasticity of the residues, we performed the scaling of each variable separately. We used the rank function in R program to reduce the amplitude of our data, meet the assumptions, and obtain the goodness of ft of the statistical models. For each variable, an independent model was created by considering each variable as the fxed efect and the others as the random efect, such that:

nooofSNPsunder selectionn~ of SNPs used +|(1 noo findividuals) oo +|(1 nofpop)(+|1n of indp− op). where, the number of SNPs under selection is the response variable; number of SNPs used is the fxed-efect variable; the terms in parentheses are the random efect variables and the number 1 indicates that the intercept is random between the observations of each variable. Te linear mixed efects model in the lme4 package does not provide the results as the coefcient of determi- nation (r²), due to theoretical problems or difculty of implementation. For this reason, we used the marginal r² to calculate the amount of variance explained by the fxed-efect variables in each model42. Finally, we used the function lm in R to establish the simple linear regression analysis of the number of SNPs used in the function of the total number of individuals. To make the graphs with the signifcant results of all the analyses, we used the ggplot2 package in R43.

Scientific Reports | (2020)10:3706 | https://doi.org/10.1038/s41598-020-60788-8 3 www.nature.com/scientificreports/ www.nature.com/scientificreports

Results Based on the 35 selected empirical articles, we analyzed 44 observations involving 36 diferent species for the number of SNPs under selection (Table 1). Tese studies were conducted mainly in Europe and North America (Fig. 1A), in ten plant families, dominance by Pinaceae (53% of the species) (Fig. 1B), distributed mainly in biomes of temperate broadleaf and mixed forest, temperate grasslands and Mediterranean forests (Fig. 1C). Te total number of individuals ranged from 22 to 2,574, and the number of SNPs used varied between 33 and 2,091,957. However, the number of SNPs under selection (SNPs with signifcant association with some geoclimatic variable) varied from 2 to 2,522 with the percentage of SNPs under selection ranging from 0.02 to 78.14 (Table 1). In the analysis using the linear mixed-efects models, we verifed the efect of each variable (related to sample size and SNP number) individually on the number of SNPs under selection, whereas the others were used as random efect, as mentioned in the methodology section. On entering, the number of populations, as fxed efect variable, the model showed no signifcant relation with the number of SNPs under selection (marginal r² = 0.011, p = 0.38). A similar result was obtained when the mean number of individuals was used as the fxed efect variable (Marginal r² = 0.041, p = 0.16). However, when the number of SNPs was used as the fxed efect variable, a positive and signifcant relationship was observed with the number of SNP under selection (Fig. 2A). When the total number of individuals was used as a fxed efect variable in the model, a negative and signifcant relation was observed in relation to the number of SNPs under selection (Fig. 2B). Subsequently, it was verifed that the total number of individuals had a negative and signifcant relation with the number of SNPs used (Fig. 2C). Discussion Te data obtained in the systematic review of studies in wild plant populations showed great heterogeneity in the number of individuals, populations, and quantity of SNPs. Our meta-analyses revealed that the total number of individuals and number of SNPs used are of fundamental importance in detecting signs of natural selection in wild plant populations. We also found an enormous knowledge gap in neotropics, underscoring the fact that predicting the response of the tropical forests to geoclimatic changes based on available data is not possible. Most of the studies had been conducted in the biomes of temperate and mixed forests of the European and North American habitats. However, we had to use them to apply the results in the recommendation for other environments, in order to help in decentralizing studies that are performed almost exclusively in the North hemisphere. Tus, it becomes evident the need to increase the possibilities of develop general methodologies guidelines that support decision making about sampling based on what we have available in the literature. Terefore, although our results were mostly based on species from temperate environment, we believe that they can be generalized to global conclusions. For this reason, our discussion focused on tropical regions, as they are recognized as a biodiversity hotspot and having really few studies with landscape genomics. We found few empirical studies evaluating SNPs under natural selection, with only 36 plants species under analysis. From those studies, we registered ten families, where 53% of the species belonged to the Pinaceae family. Tis result evidences the poor understanding of local adaptation processes for diferent species. Considering that tropical forests have approximately 11,371 species of trees44, we have a limited vision to predict how species may respond to anthropogenic changes like climate change, fragmentation, and forest loss45,46. In addition, ~81% of studies were conducted in the European continent or in North America. Although much of the plant diversity is located in tropical regions44,47, the ability of these species to adapt in situ in response to environmental changes has not been studied48. We believe that this lack of knowledge for the tropics is a refection of the socioeconomic conditions of this region as well as the lack of specialists in the area since population genomics is relatively recent and the frst studies in the tropics are just emerging (see Table 1). Considering that tropical regions could be severely afected by climate change4, our results have shown the great need for studies that seek to understand how geoclimatic variables infuence local adaptation. Trough these studies, it will be possible to predict how tropical species will respond to climate change and assist in their conservation and management29,32,35,45. With our systematic review of empirical studies, we have evidenced that the number of populations and mean number of individuals in a population do not infuence the ability to detect natural selection. Tis emphasizes the importance of sampling, at the best, the whole extent of the area that is being infuenced by the environmental variable of interest, rather than focusing on the number of populations sampled or on the mean number of individuals per population. If many populations, from similar environments, are sampled, the probability of detecting local adaptation does not increase39. On the other hand, even if few populations are sampled in contrasting environmental conditions, it would increase the power to detect local adaptation39. Tus, it will be more efective to cover the environmental heterogeneity, avoiding eforts both in the expansion of the number of populations and in the average number of individuals per population. The linear mixed-effect model showed that the number of SNPs used explained 59% of the variation in the number of SNPs under selection and therefore can positively infuence the detection of natural selection signals. Tis could occur because mostly a low percentage of genes in a genome are under natural selection (Table 1); therefore, increasing the genome sampling, also increases the chance of identifying SNPs under selection. Interestingly, a general pattern observed was that studies using Random, Targeted Pool-Seq, and Random Pool-Seq methodologies used more SNPs and had a greater number of SNPs under selection. Therefore, it would be benefcial if future works used one of these methodologies, especially Random Pool-Seq and Targeted Pool-Seq, to optimize the cost-beneft ratio. In this context, studies that aim to evaluate the indicators of natural selection in populations of wild plants could use the Targeted Pool-Seq approach, in cases where genomic information of the species is available (e.g., genome size and gene annotation). By using targeted sequencing in DNA pools, cost beneft ratio of such studies would be increased. In contrast, for the species with little or no genomic information available and for laboratories that are working on landscape genomics for the frst time, Random Pool-Seq is the best strategy to increase the chances of detecting SNPs under selection. An interesting alternative

Scientific Reports | (2020)10:3706 | https://doi.org/10.1038/s41598-020-60788-8 4 www.nature.com/scientificreports/ www.nature.com/scientificreports

Ind_ SNPs_ SNPs_ Author Species Pop Ind Pop SNPs GEA GEA* Method Eckert et al.57 Pinus taeda 54 682 12.63 1730 118 6.82 Targeted Mosca et al.58 Abies alba Mill 37 1183 31.97 249 3 1.20 Targeted Mosca et al.58 Larix decidua Mill 24 920 38.33 267 5 1.87 Targeted Keller et al.59 Populus balsamifera L 31 443 14.29 335 11 3.28 Targeted Prunier et al.60 Picea mariana 41 593 14.46 47 9 19.15 Targeted Mosca et al.58 Pinus cembra L 24 860 35.83 459 11 2.40 Targeted Mosca et al.58 Pinus mugo 27 935 34.63 693 14 2.02 Targeted Fischer et al.23 Arabidopsis halleri 5 100 20.00 2091957 1037 0.05 Random-Pool-Seq Bashalkhanov et al.31 Picea rubens 5 900 180.00 33 6 18.18 Targeted Mosca et al.61 Abies alba Mill 36 1108 30.80 231 5 2.16 Targeted Tsumura et al.62 Cryptomeria japonica 14 186 13.29 3930 25 0.64 Targeted Mosca et al.61 Larix decidua Mill 22 824 37.50 233 7 3.00 Targeted Modesto et al.63 Quercus suber L 16 96 6.00 44 5 11.36 Targeted Scalf et al.64 Picea abies [L.] Karst) 12 300 25.00 227 2 0.88 Targeted Cullingham et al.46 Pinus contorta var. latifolia 13 368 28.31 399 22 5.51 Targeted Cullingham et al.46 Pinus banksiana 4 100 25.00 399 8 2.01 Targeted Geraldes et al.65 Populus trichocarpa 25 424 16.96 28135 58 0.21 Targeted H De Kort et al.66 Frangula alnus subsp. alnus 25 619 24.76 183 143 78.14 Targeted Hamlin et al.67 Iris hexagona 8 92 11.50 750 70 9.33 Random Eckert et al.68 Pinus lambertiana 10 241 24.10 475 14 2.95 Targeted Jaramillo-Correa et al.69 Pinus pinaster Aiton 36 772 21.44 266 18 6.77 Targeted Roschanski et al.13 Abies alba Mill 4 376 94.00 267 8 3.00 Targeted Dodonaea viscosa ssp. Christmas et al.25 17 89 5.24 8462 93 1.10 Targeted Angustissima Pluess et al.10 Fagus sylvatica 79 234 2.96 144 16 11.11 Targeted Gugger et al.30 Quercus lobata 12 22 1.83 220427 79 0.04 Targeted Sork et al.20 Quercus lobata 13 45 3.46 195 8 4.10 Targeted Rellstab et al.12 Quercus robur 24 465 19.38 3576 181 5.06 Targeted-Pool-Seq Rellstab et al.12 Quercus petraea 18 350 19.44 3576 224 6.26 Targeted-Pool-Seq Rellstab et al.12 Quercus pubescens 17 326 19.18 3576 304 8.50 Targeted-Pool-Seq Di Pierro et al.18 Norway spruce 24 826 34.40 214 10 4.67 Targeted Di Pierro et al.18 Picea abies [L.] Karst 23 826 35.91 214 7 3.27 Targeted Mosca et al.19 Pinus cembra L 18 678 37.67 455 74 16.26 Targeted Mosca et al.19 Pinus mugo 20 673 33.65 663 60 9.05 Targeted Rajora et al.28 Pinus strobus 29 638 22.00 44 2 4.55 Targeted Di Pierro et al.70 Picea abies [L.] Karst 18 687 38.17 175 19 10.86 Targeted Lind et al.71 Pinus albicaulis Engelm 8 244 30.50 116231 1780 1.53 Random Fahrenkrog et al.72 Populus deltoides 50 168 3.36 79969 2.522 3.15 Targeted-Pool-Seq Frachon et al.54 Arabidopsis thaliana 168 2.574 15.32 1638649 300 0.02 Random-Pool-Seq Lanes et al.73 Ipomoea cavalcantei 1 122 122.00 34102 2239 6.57 Random Lanes et al.73 Ipomoea maurandioides 4 254 63.50 23181 1814 7.83 Random Keteleeria davidiana var. Shih et al.56 5 62 12.40 13914 15 0.11 Random-Pool-Seq formosana Martins et al.74 Quercus rugosa Née 17 103 6.06 5354 97 1.81 Random-Pool-Seq Ruiz Daniels et al.75 Pinus halepensis Mill 46 1.326 28.83 294 7 2.38 Targeted Vaccinium vitis-idaea Alam et al.76 56 56 1.00 1586 132 8.32 Random subsp. Minus Minimum 1 22 1.00 33 2 0.02 Maximum 168 2574 180 2091957 2522 78.14 Descriptive statistics Mean 25.91 520.2 29.48 97420 263.2 7 Standard deviation 27.28 471.82 32.31 395181.4 614 11.99

Table 1. Biological models and descriptive statistics calculated in 35 studies that assessed local adaptation in wild plant populations in situ using markers SNPs. (Ind, Number of individuals; Ind_Pop, Average number of individuals per population; SNPs, Number of used SNPs; Pop, Number of populations; SNPs_GEA, Number of SNPs under selection; SNPs_GEA*, Percentage of SNPs under selection; Method, Method used to generate the SNP data).

Scientific Reports | (2020)10:3706 | https://doi.org/10.1038/s41598-020-60788-8 5 www.nature.com/scientificreports/ www.nature.com/scientificreports

Figure 2. Relation between sampling design and the detection of SNPs with environmental association (A) number of SNPs with environmental association as a function of the number of SNPs used in the studies; (B) number of SNPs with environmental association as a function of the total number of individuals in each study; (C) number of SNPs used as a function of the total number of individuals in each study. Te colors of the dots represent the diferent methods used in the studies (Random used random regions of DNA and individualized sequencing; Random Pool-Seq used random regions of DNA and Pool-Seq technique; Targeted used only gene regions or expressed sequence tags and individualized sequencing; Targeted Pool-Seq used only gene regions or expressed sequence tags and sequencing in Pool-Seq).

to make landscape genomics fnancially feasible would be the use of grouped sequencing of individuals by techniques like RADseq that do not require any prior knowledge of the genome49. While it is still advantageous to use the Pool-Seq technique which can increase accuracy in allele frequency estimates, compared to individualized sequencing, making it an excellent tool to be used in landscape genomic studies50–52. Moreover, it also allows to reconcile the sampling of a large number of individuals needed in population genomics, with thousands or millions of SNPs, to evaluate adaptation in natural populations12,23,53,54. It is important to note that the Pool-Seq technique has some limitations, for example, not providing information for each individual separately52. Terefore, the use of this technique is only suitable for studies aiming for population inferences without the need of information about individuals. By fxing the variable in the linear mixed efects model as total number of individuals, we found that the total sample size was inversely related to the number of SNPs under selection and to the number of SNPs used. Although signifcant, these correlations explain only 10% of the variation in the number of SNPs under selection and 14% in the number of SNPs used. However, much of the variation in the number of SNPs under selection (59%) could be explained by fxing the variable as the number of SNPs used. We therefore, believe that the signif- icantly negative correlation between the total number of individuals and the number of SNPs under selection is a refection of the inverse relationship observed between the total number of individuals and the number of SNPs used. Tus, the negative relation seen between the total number of individuals and the number of SNPs under selection is an artifact generated by the lower capacity detection of natural selection when the number of SNPs used decreases. Te negative relationship observed between the number of individuals sampled and the number of SNPs used could be due to incomplete genomic information because of the costs associated with sequencing of whole genomes53. Among species for which some genomic information is available, size consideration of the genome becomes important; because the larger genomes may need more number of SNPs to be sampled to give a full perspective, thus impacting the number of individuals that can be analyzed given the budgetary constraints. Also, investing a huge portion of the budget for sampling in the feld to increase the sample size, would afect the laboratory stages of the study by limiting resource allocation. Tus, the negative relation found in this study, can be related to the costs of individual sequencing of each sample. In population genomics studies, the confict- ing demands between the total number of individuals and the number of SNPs used is circumvented by using the Pool-Seq technique53 (Table 1). Tis technique has also been used in plant species with incomplete or total absence of genomic information52. Tis indicates that the same strategy can be used in the landscape genomics studies of species in tropical regions, which are major biodiversity hotspots of the world44,47 with little or no genomic information available. Guidance for Study of Local Adaptation in Wild Plant Species Our study involving systematic review of empirical studies and meta-analysis of data and results therein, about wild plant populations, allows us to put forth the recommendations that in future landscape genomics studies, the experimental design should focus on increasing the total number of individuals sampled along the environmental heterogeneity under analysis39. In addition, we also suggest that researchers seek to evaluate the adaptation in wild plant populations using the Pool-Seq technique to increase the number of SNPs and accuracy in allele frequency estimates52. Tis technique has been used in population genomics as a powerful tool under diferent study scenarios, including model species54, allopolyploid species55 as well as for species with little or no genomic information available56. Following these guidelines, it will be possible to extrapolate fndings from studies that

Scientific Reports | (2020)10:3706 | https://doi.org/10.1038/s41598-020-60788-8 6 www.nature.com/scientificreports/ www.nature.com/scientificreports

are performed almost exclusively on species of temperate climate and expanding them to tropical species. In this way, landscape genomics can be conducted in tropical regions in relation to anthropogenic changes, in spite of existing budgetary constraints, and the data so generated can be used to develop strategies for management and conservation of biodiversity.

Received: 8 May 2019; Accepted: 11 February 2020; Published: xx xx xxxx

References 1. Liu, B., Su, J., Chen, J., Cui, G. & Ma, J. Anthropogenic Halo Disturbances Alter Landscape And Plant Richness: A Ripple Efect. PLoS One 8, 1–8 (2013). 2. Wilson, M. C. et al. Habitat Fragmentation And Biodiversity Conservation: Key Findings And Future Challenges. Landsc. Ecol. 31, 219–227 (2016). 3. Fahrig, L. Efects Of Habitat Fragmentation On Biodiversity. Annu. Rev. Ecol. Evol. Syst. 34, 487–515 (2003). 4. Bellard, C., Bertelsmeier, C., Leadley, P., Tuiller, W. & Courchamp, F. Impacts Of Climate Change On Te Future Of Biodiversity. Ecol. Lett. 15, 365–377 (2012). 5. Parmesan, C. & Hanley, M. E. Plants And Climate Change: Complexities And Surprises. Ann. Bot. 116, 849–864 (2015). 6. Davis, M. B., Shaw, R. G. & Etterson, J. R. Evolutionary Responses To Changing Climate. Ecology 86, 1704–1714 (2005). 7. Nicotra, A. B. et al. Plant Phenotypic Plasticity In A Changing Climate. Trends Plant Sci. 15, 684–692 (2010). 8. Alberto, F. J. et al. Potential For Evolutionary Responses To Climate Change – Evidence From Tree Populations. Glob. Chang. Biol., https://doi.org/10.1111/gcb.12181 (2013). 9. Kawecki, T. J. & Ebert, D. Conceptual Issues In Local Adaptation. Ecol. Lett. 7, 1225–1241 (2004). 10. Pluess, A. R. et al. Genome – Environment Association Study Suggests Local Adaptation To Climate At Te Regional Scale In Fagus Sylvatica. New Phytol. 210, 589–601 (2016). 11. Rellstab, C. et al. Local Adaptation (Mostly) Remains Local: Reassessing Environmental Associations Of Climate-Related Candidate Snps In Arabidopsis Halleri. Heredity (Edinb). 118, 193–201 (2017). 12. Rellstab, C. et al. Signatures Of Local Adaptation In Candidate Genes Of Oaks (Quercus Spp.) With Respect To Present And Future Climatic Conditions. Mol. Ecol. 25, 5907–5924 (2016). 13. Roschanski, A. M. et al. Evidence Of Divergent Selection For Drought And Cold Tolerance At Landscape And Local Scales In Abies Alba Mill. In Te French Mediterranean Alps. Mol. Ecol. 25, 776–794 (2016). 14. Steane, D. A. et al. Genome-Wide Scans Detect Adaptation To Aridity In A Widespread Forest Tree Species. Mol. Ecol. 23, 2500–2513 (2014). 15. Rellstab, C., Gugerli, F., Eckert, A. J., Hancock, A. M. & Holderegger, R. A Practical Guide To Environmental Association Analysis In Landscape Genomics. Mol. Ecol. 24, 4348–4370 (2015). 16. Manel, S. et al. Perspectives On Te Use Of Landscape Genetics To Detect Genetic Adaptive Variation In Te Field. Mol. Ecol. 19, 3760–3772 (2010). 17. Turner, T. L., Bourne, E. C., Von Wettberg, E. J., Hu, T. T. & Nuzhdin, S. V. Population Resequencing Reveals Local Adaptation Of Arabidopsis Lyrata To Serpentine Soils. Nat. Genet. 42, 260–263 (2010). 18. Pierro, E. A. D. et al. Climate-Related Adaptive Genetic Variation And Population Structure In Natural Stands Of Norway Spruce In Te South-Eastern Alps. Tree Genet. Genomes 12, 16, https://doi.org/10.1007/s11295-016-0972-4 (2016). 19. Mosca, E., Gugerli, F., Eckert, A. J. & Neale, D. B. Signatures Of Natural Selection On Pinus Cembra And P. Mugo Along Elevational Gradients In Te Alps. Tree Genet. Genomes, https://doi.org/10.1007/s11295-015-0964-9 (2016). 20. Sork, V. L. et al. Landscape Genomic Analysis Of Candidate Genes For Climate Adaptation In A California Endemic Oak, Quercus Lobata. Am. J. Bot. 103, 33–46 (2016). 21. Ćalić, I., Bussotti, F., Martínez-García, P. J. & Neale, D. B. Recent Landscape Genomics Studies In Forest Trees — What Can We Believe? Tree Genet. Genomes 12, 3 (2016). 22. Gaut, B. Arabidopsis Taliana As A Model For Te Genetics Of Local Adaptation. Nat. Genet. 44, 732–732 (2012). 23. Fischer, M. C. et al. Population Genomic Footprints Of Selection And Associations With Climate In Natural Populations Of Arabidopsis Halleri From Te Alps. Mol. Ecol. 22, 5594–5607 (2013). 24. Andrews, K. R. & Luikart, G. Recent Novel Approaches For Population Genomics Data Analysis. Mol. Ecol. 23, 1661–1667 (2014). 25. Christmas, M. J., Bifn, E., Breed, M. F. & Lowe, A. J. Finding Needles In A Genomic Haystack: Targeted Capture Identifes Clear Signatures Of Selection In A Nonmodel Plant Species. Mol. Ecol. 25, 4216–4233 (2016). 26. Parisod, C. & Holderegger, R. Adaptive Landscape Genetics: Pitfalls And Benefts Adaptive Landscape Genetics: Pitfalls. Mol. Biol. Evol. 21, 3644–3646 (2012). 27. Zhou, Y., Zhang, L., Liu, J., Wu, G. & Savolainen, O. Climatic Adaptation And Ecological Divergence Between Two Closely Related Pine Species In Southeast China. Mol. Ecol. 23, 3504–3522 (2014). 28. Rajora, O. P., Eckert, A. J. & Zinck, J. W. R. Single-Locus Versus Multilocus Patterns Of Local Adaptation To Climate In Eastern White Pine (Pinus Strobus, Pinaceae). PLoS One 11, 1–26 (2016). 29. Fitzpatrick, M. C. & Keller, S. Ecological Genomics Meets Community-Level Modelling Of Biodiversity: Mapping Te Genomic Landscape Of Current And Future Environmental. Ecol. Lett. 18, 1–16 (2015). 30. Gugger, P. F., Cokus, S. J. & Sork, V. L. Association Of Transcriptome-Wide Sequence Variation With Climate Gradients In Valley Oak (Quercus Lobata). Tree Genet. Genomes 12, 1–14 (2016). 31. Bashalkhanov, S., Eckert, A. J. & Rajora, O. P. Genetic Signatures Of Natural Selection In Response To Air Pollution In Red Spruce (Picea Rubens, Pinaceae). Mol. Ecol. 22, 5877–5889 (2013). 32. Shafer, A. B. et al. Genomics And Te Challenging Translation Into Conservation Practice. Trends Ecol. Evol. 30, 78–87 (2015). 33. Schoville, S. D. et al. Adaptive Genetic Variation On Te Landscape: Methods And Cases Adaptive Genetic Variation On Te Landscape: Methods And Cases. Annu. Rev. Ecol. Evol. Syst. 43, 23–43 (2012). 34. Robasky, K., Lewis, N. E. & Church, G. M. Te Role Of Replicates For Error Mitigation In Next-Generation Sequencing. Nat. Rev. Genet. 15, 56 (2014). 35. Benestan, L. M. et al. Conservation Genomics Of Natural And Managed Populations: Building A Conceptual And Practical Framework. Mol. Ecol. 25, 2967–2977 (2016). 36. Manel, S. et al. Genomic Resources And Teir Infuence On Te Detection Of Te Signal Of Positive Selection In Genome Scans. Mol. Ecol. 25, 170–184 (2016). 37. Ahrens, C. W. et al. Te Search For Loci Under Selection: Trends, Biases And Progress. Mol. Ecol. https://doi.org/10.1111/mec.14549 (2018). 38. Mita, S. D. E. et al. Detecting Selection Along Environmental Gradients: Analysis Of Eight Methods And Teir Efectiveness For Outbreeding And Selfng Populations. Mol. Ecol. 22, 1383–1399 (2013). 39. Lotterhos, K. E. & Whitlock, M. C. Te Relative Power Of Genome Scans To Detect Local Adaptation Depends On Sampling Design And Statistical Method. Mol. Ecol. 24, 1031–1046 (2015).

Scientific Reports | (2020)10:3706 | https://doi.org/10.1038/s41598-020-60788-8 7 www.nature.com/scientificreports/ www.nature.com/scientificreports

40. Forester, B. R., Jones, M. R., Joost, S., Landguth, E. L. & Lasky, J. R. Detecting Selection In Natural Populations: Making Sense Of Detecting Spatial Genetic Signatures Of Local Adaptation In Heterogeneous Landscapes. Mol. Ecol. 25, 104–120 (2016). 41. Bates, D., Mächler, M., Bolker, B. M. & Walker, S. C. Fitting Linear Mixed-Efects Models Using Lme4. J. Stat. Sofw. Oct. 67 (2015). 42. Nakagawa, S. & Schielzeth, H. A General And Simple Method For Obtaining R 2 From Generalized Linear Mixed-Efects Models. Methods Ecol. and Evolution 4, 133–142 (2013). 43. H, W. ggplot2: Elegant Graphics For Data Analysis. Springer (2016). 44. Slik, J. W. F. et al. An Estimate Of Te Number Of Tropical Tree Species. Pnas 112, E4628–E4629 (2015). 45. Hansen, M. M., Olivieri, I., Waller, D. M. & Nielsen, E. E. Monitoring Adaptive Genetic Responses To Environmental Change. Mol. Ecol. 21, 1311–1329 (2012). 46. Cullingham, C. I., Cooke, J. E. K. & Coltman, D. W. Cross-Species Outlier Detection Reveals Diferent Evolutionary Pressures Between Sister Species. New Phytol. 204, 215–229 (2014). 47. Mannion, P. D., Upchurch, P., Benson, R. B. J. & Goswami, A. Te Latitudinal Biodiversity Gradient Trough Deep Time. Trends Ecol. Evol. 29, 42–50 (2014). 48. Storfer, A., Murphy, M. A., Spear, S. F., Holderegger, R. & Waits, L. P. Landscape Genetics: Where Are We Now? Mol. Ecol. 19, 3496–3514 (2010). 49. Andrews, K. R., Good, J. M., Miller, M. R., Luikart, G. & Hohenlohe, P. A. Harnessing Te Power Of Radseq For Ecological And Evolutionary Genomics. Nat. Rev. Genet. 17, 81–92 (2016). 50. Futschik, A. & Schlötterer, C. Te Next Generation Of Molecular Markers From Massively Parallel Sequencing Of Pooled DNA Samples. Genetics 186, 207–218 (2010). 51. Bansal, V., Tewhey, R., Leproust, E. M. & Schork, N. J. Efcient And Cost Efective Population Resequencing By Pooling And In- Solution Hybridization. PLoS One 6, 1–6 (2011). 52. Schlötterer, C., Tobler, R., Kofer, R. & Nolte, V. Sequencing Pools Of Individuals — Mining Genome-Wide Polymorphism Data Without Big Funding. Nat. Rev. Genet. 15, 749 (2014). 53. Gautier, M. et al. Estimation Of Population Allele Frequencies From Next-Generation Sequencing Data: Pool-Versus Individual- Based Genotyping. Mol. Ecol. 22, 3766–3779 (2013). 54. Frachon, L. et al. A Genomic Map Of Climate Adaptation In Arabidopsis Taliana At A Micro-Geographic Scale. Front. Plant Sci. 9, 1–15 (2018). 55. Hirao, A. S. et al. Cost-Efective Discovery Of Nucleotide Polymorphisms In Populations Of An Allopolyploid Species Using Pool- Seq. Am. J. Mol. Biol. 7, 153–168 (2017). 56. Shih, K., Chang, C., Chung, J., Chiang, Y. & Hwang, S.-Y. Adaptive Genetic Divergence Despite Signifcant Isolation-By-Distance In Populations Of Taiwan Cow-Tail Fir (Keteleeria Davidiana Var. Formosana). Front. Plant Sci. 9 (2018). 57. Eckert, A. J. et al. Back To Nature: Ecological Genomics Of Loblolly Pine (Pinus Taeda, Pinaceae). Mol. Ecol. 19, 3789–3805 (2010). 58. Mosca, E. et al. The Geographical And Environmental Determinants Of Genetic Diversity For Four Alpine Conifers Of The European Alps. Mol. Ecol. 21, 5530–5545 (2012). 59. Keller, S. R., Levsen, N., Olson, M. S. & Tifn, P. Local Adaptation in the Flowering-Time Gene Network of Balsam Poplar, Populus balsamifera L. Mol. Biol. Evol. 29, 3143–3152 (2012). 60. Prunier, J., GÉrardi, S., Laroche, J., Beaulieu, J. & Bousquet, J. Parallel And Lineage-Specifc Molecular Adaptation To Climate In Boreal Black Spruce. Mol. Ecol. 21, 4270–4286 (2012). 61. Mosca, E., González-Martíınez, S. C. & Neale, D. B. Environmental Versus Geographical Determinants Of Genetic Structure In Two Subalpine Conifers. New Phytol. 201, 180–192 (2014). 62. Tsumura, Y. et al. Genetic Diferentiation And Evolutionary Adaptation In Cryptomeria Japonica. G3 Genes|Genomes|Genetics 4, 2389–402 (2014). 63. Modesto, I. S. et al. Identifying Signatures Of Natural Selection In Cork Oak (Quercus Suber L.) Genes Trough SNP Analysis. Tree Genet. Genomes 10, 1645–1660 (2014). 64. Scalf, M. et al. Micro- And Macro-Geographic Scale Efect On Te Molecular Imprint Of Selection And Adaptation In Norway Spruce. PLoS Genet. 9, 1–22 (2014). 65. Geraldes, A. et al. Landscape Genomics Of Populus Trichocarpa: Te Role Of Hybridization, Limited Gene Flow, And Natural Selection In Shaping Patterns Of Population Structure. Evolution (N. Y). 68, 3260–3280 (2014). 66. Kort, H. D., Vandepitte, K., Mergeay, J., Mijnsbrugge, K. V. & Honnay, O. Te Population Genomic Signature Of Environmental Selection In Te Widespread Insect-Pollinated Tree Species Frangula Alnus At Diferent Geographical Scales. Heredity (Edinb). 115, 415–425 (2015). 67. Hamlin, J. A. P. & Arnold, M. L. Neutral And Selective Processes Drive Population Diferentiation For Iris Hexagona. J. Hered. 628–636, https://doi.org/10.5061/dryad.4n15b (2015). 68. Eckert, A. J. et al. Local Adaptation At Fine Spatial Scales: An Example From Sugar Pine (Pinus Lambertiana, Pinaceae). Tree Genet. Genomes 11 (2015). 69. Jaramillo-Correa, J.-P. et al. Molecular Proxies For Climate Maladaptation In A Long-Lived Tree (Pinus Pinaster Aiton, Pinaceae). Genetics 199, 793–807 (2015). 70. Pierro, E. A. D. et al. Adaptive Variation In Natural Alpine Populations Of Norway Spruce (Picea Abies [L.] Karst) At Regional Scale: Landscape Features And Altitudinal Gradient Efects. For. Ecol. Manage. 405, 350–359 (2017). 71. Lind, B. M. et al. Water Availability Drives Signatures Of Local Adaptation In Whitebark Pine (Pinus Albicaulis Engelm.) Across Fine Spatial Scales Of Te Lake Tahoe Basin, USA. Mol. Ecol. 26, 3168–3185 (2017). 72. Fahrenkrog, A. M. et al. Population Genomics Of Te Eastern Cottonwood (Populus Deltoides). Ecol. Evol. 9426–9440, https://doi. org/10.1002/ece3.3466 (2017). 73. Lanes, É. C. et al. Landscape Genomic Conservation Assessment Of A Narrow-Endemic And A Widespread Morning Glory From Amazonian Savannas. Front. Genet. 9, 1–13 (2018). 74. Martins, K. et al. Landscape Genomics Provides Evidence Of Climate-Associated Genetic Variation In Mexican Populations Of Quercus Rugosa. Evol. Appl. 11, 1842–1858 (2018). 75. Daniels, R. R. et al. Inferring Selection In Instances Of Long ‐ Range Colonization: Te Aleppo Pine (Pinus Halepensis) In Te Mediterranean Basin. Mol. Ecol. 27, 3331–3345 (2018). 76. Alam, Z., Roncal, J. & Peña-Castillo, L. Genetic Variation Associated With Healthy Traits And Environmental Conditions In Vaccinium Vitis-Idaea. BMC Genomics 19, 1–13 (2018).

Acknowledgements Te authors thank the Brazilian National Council of Scientifc and Technological Development (CNPq) for the FAG research productivity fellowship; and Coordination for the Improvement of Higher Education Personnel (CAPES) for granting a scholarship to ASS. Te authors also thank Marina Corrêa Côrtes, Luciana Aparecida Carlini Garcia, and Ricardo Dobrovolski for the corrections in the frst version of the manuscript. Te authors acknowledge Dr. Leandro L. Loguercio and Dra. Thâmara M. Lima who also helped on the writing of the resubmitting letter.

Scientific Reports | (2020)10:3706 | https://doi.org/10.1038/s41598-020-60788-8 8 www.nature.com/scientificreports/ www.nature.com/scientificreports

Author contributions A.S.S. and F.A.G. contributed to the design, analysis and writing of the manuscript. Competing interests Te authors declare no competing interests. Additional information Correspondence and requests for materials should be addressed to A.S.S. Reprints and permissions information is available at www.nature.com/reprints. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional afliations. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre- ative Commons license, and indicate if changes were made. Te images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Scientific Reports | (2020)10:3706 | https://doi.org/10.1038/s41598-020-60788-8 9